bsd/netinet/udp_usrreq.c

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22 /*
  23  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
  24  *      The Regents of the University of California.  All rights reserved.
  25  *
  26  * Redistribution and use in source and binary forms, with or without
  27  * modification, are permitted provided that the following conditions
  28  * are met:
  29  * 1. Redistributions of source code must retain the above copyright
  30  *    notice, this list of conditions and the following disclaimer.
  31  * 2. Redistributions in binary form must reproduce the above copyright
  32  *    notice, this list of conditions and the following disclaimer in the
  33  *    documentation and/or other materials provided with the distribution.
  34  * 3. All advertising materials mentioning features or use of this software
  35  *    must display the following acknowledgement:
  36  *      This product includes software developed by the University of
  37  *      California, Berkeley and its contributors.
  38  * 4. Neither the name of the University nor the names of its contributors
  39  *    may be used to endorse or promote products derived from this software
  40  *    without specific prior written permission.
  41  *
  42  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  43  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  44  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  45  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  46  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  47  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  48  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  49  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  50  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  51  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  52  * SUCH DAMAGE.
  53  *
  54  *      @(#)udp_usrreq.c        8.6 (Berkeley) 5/23/95
  55  * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.13 2001/08/08 18:59:54 ghelmer Exp $
  56  */
  57
  58 #include <sys/param.h>
  59 #include <sys/systm.h>
  60 #include <sys/kernel.h>
  61 #include <sys/malloc.h>
  62 #include <sys/mbuf.h>
  63 #include <sys/domain.h>
  64 #include <sys/protosw.h>
  65 #include <sys/socket.h>
  66 #include <sys/socketvar.h>
  67 #include <sys/sysctl.h>
  68 #include <sys/syslog.h>
  69
  70 #include <net/if.h>
  71 #include <net/route.h>
  72
  73 #include <netinet/in.h>
  74 #include <netinet/in_systm.h>
  75 #include <netinet/ip.h>
  76 #if INET6
  77 #include <netinet/ip6.h>
  78 #endif
  79 #include <netinet/in_pcb.h>
  80 #include <netinet/in_var.h>
  81 #include <netinet/ip_var.h>
  82 #if INET6
  83 #include <netinet6/ip6_var.h>
  84 #endif
  85 #include <netinet/ip_icmp.h>
  86 #include <netinet/icmp_var.h>
  87 #include <netinet/udp.h>
  88 #include <netinet/udp_var.h>
  89 #include <sys/kdebug.h>
  90
  91 #if IPSEC
  92 #include <netinet6/ipsec.h>
  93 extern int ipsec_bypass;
  94 #endif /*IPSEC*/
  95
  96
  97 #define DBG_LAYER_IN_BEG        NETDBG_CODE(DBG_NETUDP, 0)
  98 #define DBG_LAYER_IN_END        NETDBG_CODE(DBG_NETUDP, 2)
  99 #define DBG_LAYER_OUT_BEG       NETDBG_CODE(DBG_NETUDP, 1)
 100 #define DBG_LAYER_OUT_END       NETDBG_CODE(DBG_NETUDP, 3)
 101 #define DBG_FNC_UDP_INPUT       NETDBG_CODE(DBG_NETUDP, (5 << 8))
 102 #define DBG_FNC_UDP_OUTPUT      NETDBG_CODE(DBG_NETUDP, (6 << 8) | 1)
 103
 104
 105 #define __STDC__ 1
 106 /*
 107  * UDP protocol implementation.
 108  * Per RFC 768, August, 1980.
 109  */
 110 #ifndef COMPAT_42
 111 static int      udpcksum = 1;
 112 #else
 113 static int      udpcksum = 0;           /* XXX */
 114 #endif
 115 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
 116                 &udpcksum, 0, "");
 117
 118 int     log_in_vain = 0;
 119 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
 120     &log_in_vain, 0, "Log all incoming UDP packets");
 121
 122 static int      blackhole = 0;
 123 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
 124         &blackhole, 0, "Do not send port unreachables for refused connects");
 125
 126 struct  inpcbhead udb;          /* from udp_var.h */
 127 #define udb6    udb  /* for KAME src sync over BSD*'s */
 128 struct  inpcbinfo udbinfo;
 129
 130 #ifndef UDBHASHSIZE
 131 #define UDBHASHSIZE 16
 132 #endif
 133
 134 extern  int apple_hwcksum_rx;
 135
 136 struct  udpstat udpstat;        /* from udp_var.h */
 137 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RD,
 138     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
 139
 140 static struct   sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
 141 #if INET6
 142 struct udp_in6 {
 143         struct sockaddr_in6     uin6_sin;
 144         u_char                  uin6_init_done : 1;
 145 } udp_in6 = {
 146         { sizeof(udp_in6.uin6_sin), AF_INET6 },
 147         0
 148 };
 149 struct udp_ip6 {
 150         struct ip6_hdr          uip6_ip6;
 151         u_char                  uip6_init_done : 1;
 152 } udp_ip6;
 153 #endif /* INET6 */
 154
 155 static void udp_append __P((struct inpcb *last, struct ip *ip,
 156                             struct mbuf *n, int off));
 157 #if INET6
 158 static void ip_2_ip6_hdr __P((struct ip6_hdr *ip6, struct ip *ip));
 159 #endif
 160
 161 static int udp_detach __P((struct socket *so));
 162 static  int udp_output __P((struct inpcb *, struct mbuf *, struct sockaddr *,
 163                             struct mbuf *, struct proc *));
 164
 165 void
 166 udp_init()
 167 {
 168         vm_size_t       str_size;
 169         int             stat;
 170         u_char          fake_owner;
 171         struct in_addr  laddr;
 172         struct in_addr  faddr;
 173         u_short         lport;
 174
 175         LIST_INIT(&udb);
 176         udbinfo.listhead = &udb;
 177         udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
 178         udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
 179                                         &udbinfo.porthashmask);
 180 #ifdef __APPLE__
 181         str_size = (vm_size_t) sizeof(struct inpcb);
 182         udbinfo.ipi_zone = (void *) zinit(str_size, 80000*str_size, 8192, "udpcb");
 183
 184         udbinfo.last_pcb = 0;
 185         in_pcb_nat_init(&udbinfo, AF_INET, IPPROTO_UDP, SOCK_DGRAM);
 186 #else
 187         udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets,
 188                                  ZONE_INTERRUPT, 0);
 189 #endif
 190
 191 #if 0
 192         /* for pcb sharing testing only */
 193         stat = in_pcb_new_share_client(&udbinfo, &fake_owner);
 194         kprintf("udp_init in_pcb_new_share_client - stat = %d\n", stat);
 195
 196         laddr.s_addr = 0x11646464;
 197         faddr.s_addr = 0x11646465;
 198
 199         lport = 1500;
 200         in_pcb_grab_port(&udbinfo, 0, laddr, &lport, faddr, 1600, 0, fake_owner);
 201         kprintf("udp_init in_pcb_grab_port - stat = %d\n", stat);
 202
 203         stat = in_pcb_rem_share_client(&udbinfo, fake_owner);
 204         kprintf("udp_init in_pcb_rem_share_client - stat = %d\n", stat);
 205
 206         stat = in_pcb_new_share_client(&udbinfo, &fake_owner);
 207         kprintf("udp_init in_pcb_new_share_client(2) - stat = %d\n", stat);
 208
 209         laddr.s_addr = 0x11646464;
 210         faddr.s_addr = 0x11646465;
 211
 212         lport = 1500;
 213         stat = in_pcb_grab_port(&udbinfo, 0, laddr, &lport, faddr, 1600, 0, fake_owner);
 214         kprintf("udp_init in_pcb_grab_port(2) - stat = %d\n", stat);
 215 #endif
 216 }
 217
 218 void
 219 udp_input(m, iphlen)
 220         register struct mbuf *m;
 221         int iphlen;
 222 {
 223         register struct ip *ip;
 224         register struct udphdr *uh;
 225         register struct inpcb *inp;
 226         struct mbuf *opts = 0;
 227         int len;
 228         struct ip save_ip;
 229         struct sockaddr *append_sa;
 230
 231         udpstat.udps_ipackets++;
 232
 233         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_START, 0,0,0,0,0);
 234         if (m->m_pkthdr.csum_flags & CSUM_TCP_SUM16)
 235                 m->m_pkthdr.csum_flags = 0; /* invalidate hwcksum for UDP */
 236
 237         /*
 238          * Strip IP options, if any; should skip this,
 239          * make available to user, and use on returned packets,
 240          * but we don't yet have a way to check the checksum
 241          * with options still present.
 242          */
 243         if (iphlen > sizeof (struct ip)) {
 244                 ip_stripoptions(m, (struct mbuf *)0);
 245                 iphlen = sizeof(struct ip);
 246         }
 247
 248         /*
 249          * Get IP and UDP header together in first mbuf.
 250          */
 251         ip = mtod(m, struct ip *);
 252         if (m->m_len < iphlen + sizeof(struct udphdr)) {
 253                 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
 254                         udpstat.udps_hdrops++;
 255                         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 256                         return;
 257                 }
 258                 ip = mtod(m, struct ip *);
 259         }
 260         uh = (struct udphdr *)((caddr_t)ip + iphlen);
 261
 262         /* destination port of 0 is illegal, based on RFC768. */
 263         if (uh->uh_dport == 0)
 264                 goto bad;
 265
 266         KERNEL_DEBUG(DBG_LAYER_IN_BEG, uh->uh_dport, uh->uh_sport,
 267                      ip->ip_src.s_addr, ip->ip_dst.s_addr, uh->uh_ulen);
 268
 269         /*
 270          * Make mbuf data length reflect UDP length.
 271          * If not enough data to reflect UDP length, drop.
 272          */
 273         len = ntohs((u_short)uh->uh_ulen);
 274         if (ip->ip_len != len) {
 275                 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
 276                         udpstat.udps_badlen++;
 277                         goto bad;
 278                 }
 279                 m_adj(m, len - ip->ip_len);
 280                 /* ip->ip_len = len; */
 281         }
 282         /*
 283          * Save a copy of the IP header in case we want restore it
 284          * for sending an ICMP error message in response.
 285          */
 286         save_ip = *ip;
 287
 288         /*
 289          * Checksum extended UDP header and data.
 290          */
 291         if (uh->uh_sum) {
 292                if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
 293                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
 294                                 uh->uh_sum = m->m_pkthdr.csum_data;
 295                         else
 296                             goto doudpcksum;
 297                       uh->uh_sum ^= 0xffff;
 298                 } else {
 299 doudpcksum:
 300                         bzero(((struct ipovly *)ip)->ih_x1, 9);
 301                         ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
 302                         uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
 303                 }
 304                 if (uh->uh_sum) {
 305                         udpstat.udps_badsum++;
 306                         m_freem(m);
 307                         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 308                         return;
 309                 }
 310         }
 311 #ifndef __APPLE__
 312          else
 313                 udpstat.udps_nosum++;
 314 #endif
 315
 316         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
 317             in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
 318                 struct inpcb *last;
 319                 /*
 320                  * Deliver a multicast or broadcast datagram to *all* sockets
 321                  * for which the local and remote addresses and ports match
 322                  * those of the incoming datagram.  This allows more than
 323                  * one process to receive multi/broadcasts on the same port.
 324                  * (This really ought to be done for unicast datagrams as
 325                  * well, but that would cause problems with existing
 326                  * applications that open both address-specific sockets and
 327                  * a wildcard socket listening to the same port -- they would
 328                  * end up receiving duplicates of every unicast datagram.
 329                  * Those applications open the multiple sockets to overcome an
 330                  * inadequacy of the UDP socket interface, but for backwards
 331                  * compatibility we avoid the problem here rather than
 332                  * fixing the interface.  Maybe 4.5BSD will remedy this?)
 333                  */
 334
 335                 /*
 336                  * Construct sockaddr format source address.
 337                  */
 338                 udp_in.sin_port = uh->uh_sport;
 339                 udp_in.sin_addr = ip->ip_src;
 340                 /*
 341                  * Locate pcb(s) for datagram.
 342                  * (Algorithm copied from raw_intr().)
 343                  */
 344                 last = NULL;
 345 #if INET6
 346                 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
 347 #endif
 348                 LIST_FOREACH(inp, &udb, inp_list) {
 349 #ifdef __APPLE__
 350                         /* Ignore nat/SharedIP dummy pcbs */
 351                         if (inp->inp_socket == &udbinfo.nat_dummy_socket)
 352                                 continue;
 353 #endif
 354 #if INET6
 355                         if ((inp->inp_vflag & INP_IPV4) == 0)
 356                                 continue;
 357 #endif
 358                         if (inp->inp_lport != uh->uh_dport)
 359                                 continue;
 360                         if (inp->inp_laddr.s_addr != INADDR_ANY) {
 361                                 if (inp->inp_laddr.s_addr !=
 362                                     ip->ip_dst.s_addr)
 363                                         continue;
 364                         }
 365                         if (inp->inp_faddr.s_addr != INADDR_ANY) {
 366                                 if (inp->inp_faddr.s_addr !=
 367                                     ip->ip_src.s_addr ||
 368                                     inp->inp_fport != uh->uh_sport)
 369                                         continue;
 370                         }
 371
 372                         if (last != NULL) {
 373                                 struct mbuf *n;
 374
 375 #if IPSEC
 376                                 /* check AH/ESP integrity. */
 377                                 if (ipsec_bypass == 0 && ipsec4_in_reject_so(m, last->inp_socket)) {
 378                                         ipsecstat.in_polvio++;
 379                                         /* do not inject data to pcb */
 380                                 } else
 381 #endif /*IPSEC*/
 382                                 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
 383                                         udp_append(last, ip, n,
 384                                                    iphlen +
 385                                                    sizeof(struct udphdr));
 386                                 }
 387                         }
 388                         last = inp;
 389                         /*
 390                          * Don't look for additional matches if this one does
 391                          * not have either the SO_REUSEPORT or SO_REUSEADDR
 392                          * socket options set.  This heuristic avoids searching
 393                          * through all pcbs in the common case of a non-shared
 394                          * port.  It * assumes that an application will never
 395                          * clear these options after setting them.
 396                          */
 397                         if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
 398                                 break;
 399                 }
 400
 401                 if (last == NULL) {
 402                         /*
 403                          * No matching pcb found; discard datagram.
 404                          * (No need to send an ICMP Port Unreachable
 405                          * for a broadcast or multicast datgram.)
 406                          */
 407                         udpstat.udps_noportbcast++;
 408                         goto bad;
 409                 }
 410 #if IPSEC
 411                 /* check AH/ESP integrity. */
 412                 if (ipsec_bypass == 0 && m && ipsec4_in_reject_so(m, last->inp_socket)) {
 413                         ipsecstat.in_polvio++;
 414                         goto bad;
 415                 }
 416 #endif /*IPSEC*/
 417                 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
 418                 return;
 419         }
 420         /*
 421          * Locate pcb for datagram.
 422          */
 423         inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
 424             ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
 425         if (inp == NULL) {
 426                 if (log_in_vain) {
 427                         char buf[4*sizeof "123"];
 428
 429                         strcpy(buf, inet_ntoa(ip->ip_dst));
 430                         log(LOG_INFO,
 431                             "Connection attempt to UDP %s:%d from %s:%d\n",
 432                             buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
 433                             ntohs(uh->uh_sport));
 434                 }
 435                 udpstat.udps_noport++;
 436                 if (m->m_flags & (M_BCAST | M_MCAST)) {
 437                         udpstat.udps_noportbcast++;
 438                         goto bad;
 439                 }
 440 #if ICMP_BANDLIM
 441                 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
 442                         goto bad;
 443 #endif
 444                 if (blackhole)
 445                         goto bad;
 446                 *ip = save_ip;
 447                 ip->ip_len += iphlen;
 448                 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
 449                 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 450                 return;
 451         }
 452 #if IPSEC
 453         if (ipsec_bypass == 0 && inp != NULL && ipsec4_in_reject_so(m, inp->inp_socket)) {
 454                 ipsecstat.in_polvio++;
 455                 goto bad;
 456         }
 457 #endif /*IPSEC*/
 458
 459         /*
 460          * Construct sockaddr format source address.
 461          * Stuff source address and datagram in user buffer.
 462          */
 463         udp_in.sin_port = uh->uh_sport;
 464         udp_in.sin_addr = ip->ip_src;
 465         if (inp->inp_flags & INP_CONTROLOPTS
 466             || inp->inp_socket->so_options & SO_TIMESTAMP) {
 467 #if INET6
 468                 if (inp->inp_vflag & INP_IPV6) {
 469                         int savedflags;
 470
 471                         ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
 472                         savedflags = inp->inp_flags;
 473                         inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
 474                         ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
 475                         inp->inp_flags = savedflags;
 476                 } else
 477 #endif
 478                 ip_savecontrol(inp, &opts, ip, m);
 479         }
 480         m_adj(m, iphlen + sizeof(struct udphdr));
 481
 482         KERNEL_DEBUG(DBG_LAYER_IN_END, uh->uh_dport, uh->uh_sport,
 483                      save_ip.ip_src.s_addr, save_ip.ip_dst.s_addr, uh->uh_ulen);
 484
 485 #if INET6
 486         if (inp->inp_vflag & INP_IPV6) {
 487                 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
 488                 append_sa = (struct sockaddr *)&udp_in6;
 489         } else
 490 #endif
 491         append_sa = (struct sockaddr *)&udp_in;
 492         if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
 493                 udpstat.udps_fullsock++;
 494                 goto bad;
 495         }
 496         sorwakeup(inp->inp_socket);
 497         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 498         return;
 499 bad:
 500         m_freem(m);
 501         if (opts)
 502                 m_freem(opts);
 503         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 504         return;
 505 }
 506
 507 #if INET6
 508 static void
 509 ip_2_ip6_hdr(ip6, ip)
 510         struct ip6_hdr *ip6;
 511         struct ip *ip;
 512 {
 513         bzero(ip6, sizeof(*ip6));
 514
 515         ip6->ip6_vfc = IPV6_VERSION;
 516         ip6->ip6_plen = ip->ip_len;
 517         ip6->ip6_nxt = ip->ip_p;
 518         ip6->ip6_hlim = ip->ip_ttl;
 519         ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
 520                 IPV6_ADDR_INT32_SMP;
 521         ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
 522         ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
 523 }
 524 #endif
 525
 526 /*
 527  * subroutine of udp_input(), mainly for source code readability.
 528  * caller must properly init udp_ip6 and udp_in6 beforehand.
 529  */
 530 static void
 531 udp_append(last, ip, n, off)
 532         struct inpcb *last;
 533         struct ip *ip;
 534         struct mbuf *n;
 535         int off;
 536 {
 537         struct sockaddr *append_sa;
 538         struct mbuf *opts = 0;
 539
 540         if (last->inp_flags & INP_CONTROLOPTS ||
 541             last->inp_socket->so_options & SO_TIMESTAMP) {
 542 #if INET6
 543                 if (last->inp_vflag & INP_IPV6) {
 544                         int savedflags;
 545
 546                         if (udp_ip6.uip6_init_done == 0) {
 547                                 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
 548                                 udp_ip6.uip6_init_done = 1;
 549                         }
 550                         savedflags = last->inp_flags;
 551                         last->inp_flags &= ~INP_UNMAPPABLEOPTS;
 552                         ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
 553                         last->inp_flags = savedflags;
 554                 } else
 555 #endif
 556                 ip_savecontrol(last, &opts, ip, n);
 557         }
 558 #if INET6
 559         if (last->inp_vflag & INP_IPV6) {
 560                 if (udp_in6.uin6_init_done == 0) {
 561                         in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
 562                         udp_in6.uin6_init_done = 1;
 563                 }
 564                 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
 565         } else
 566 #endif
 567         append_sa = (struct sockaddr *)&udp_in;
 568         m_adj(n, off);
 569         if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
 570                 m_freem(n);
 571                 if (opts)
 572                         m_freem(opts);
 573                 udpstat.udps_fullsock++;
 574         } else
 575                 sorwakeup(last->inp_socket);
 576 }
 577
 578 /*
 579  * Notify a udp user of an asynchronous error;
 580  * just wake up so that he can collect error status.
 581  */
 582 void
 583 udp_notify(inp, errno)
 584         register struct inpcb *inp;
 585         int errno;
 586 {
 587         inp->inp_socket->so_error = errno;
 588         sorwakeup(inp->inp_socket);
 589         sowwakeup(inp->inp_socket);
 590 }
 591
 592 void
 593 udp_ctlinput(cmd, sa, vip)
 594         int cmd;
 595         struct sockaddr *sa;
 596         void *vip;
 597 {
 598         struct ip *ip = vip;
 599         struct udphdr *uh;
 600         void (*notify) __P((struct inpcb *, int)) = udp_notify;
 601         struct in_addr faddr;
 602         struct inpcb *inp;
 603         int s;
 604
 605         faddr = ((struct sockaddr_in *)sa)->sin_addr;
 606         if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
 607                 return;
 608
 609         if (PRC_IS_REDIRECT(cmd)) {
 610                 ip = 0;
 611                 notify = in_rtchange;
 612         } else if (cmd == PRC_HOSTDEAD)
 613                 ip = 0;
 614         else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
 615                 return;
 616         if (ip) {
 617                 s = splnet();
 618                 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
 619                 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
 620                     ip->ip_src, uh->uh_sport, 0, NULL);
 621                 if (inp != NULL && inp->inp_socket != NULL)
 622                         (*notify)(inp, inetctlerrmap[cmd]);
 623                 splx(s);
 624         } else
 625                 in_pcbnotifyall(&udb, faddr, inetctlerrmap[cmd], notify);
 626 }
 627
 628 static int
 629 udp_pcblist SYSCTL_HANDLER_ARGS
 630 {
 631         int error, i, n, s;
 632         struct inpcb *inp, **inp_list;
 633         inp_gen_t gencnt;
 634         struct xinpgen xig;
 635
 636         /*
 637          * The process of preparing the TCB list is too time-consuming and
 638          * resource-intensive to repeat twice on every request.
 639          */
 640         if (req->oldptr == 0) {
 641                 n = udbinfo.ipi_count;
 642                 req->oldidx = 2 * (sizeof xig)
 643                         + (n + n/8) * sizeof(struct xinpcb);
 644                 return 0;
 645         }
 646
 647         if (req->newptr != 0)
 648                 return EPERM;
 649
 650         /*
 651          * OK, now we're committed to doing something.
 652          */
 653         s = splnet();
 654         gencnt = udbinfo.ipi_gencnt;
 655         n = udbinfo.ipi_count;
 656         splx(s);
 657
 658         xig.xig_len = sizeof xig;
 659         xig.xig_count = n;
 660         xig.xig_gen = gencnt;
 661         xig.xig_sogen = so_gencnt;
 662         error = SYSCTL_OUT(req, &xig, sizeof xig);
 663         if (error)
 664                 return error;
 665     /*
 666      * We are done if there is no pcb
 667      */
 668     if (n == 0)
 669         return 0;
 670
 671         inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
 672         if (inp_list == 0) {
 673                 return ENOMEM;
 674         }
 675
 676         for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
 677              inp = LIST_NEXT(inp, inp_list)) {
 678                 if (inp->inp_gencnt <= gencnt)
 679                         inp_list[i++] = inp;
 680         }
 681         splx(s);
 682         n = i;
 683
 684         error = 0;
 685         for (i = 0; i < n; i++) {
 686                 inp = inp_list[i];
 687                 if (inp->inp_gencnt <= gencnt) {
 688                         struct xinpcb xi;
 689                         xi.xi_len = sizeof xi;
 690                         /* XXX should avoid extra copy */
 691                         bcopy(inp, &xi.xi_inp, sizeof *inp);
 692                         if (inp->inp_socket)
 693                                 sotoxsocket(inp->inp_socket, &xi.xi_socket);
 694                         error = SYSCTL_OUT(req, &xi, sizeof xi);
 695                 }
 696         }
 697         if (!error) {
 698                 /*
 699                  * Give the user an updated idea of our state.
 700                  * If the generation differs from what we told
 701                  * her before, she knows that something happened
 702                  * while we were processing this request, and it
 703                  * might be necessary to retry.
 704                  */
 705                 s = splnet();
 706                 xig.xig_gen = udbinfo.ipi_gencnt;
 707                 xig.xig_sogen = so_gencnt;
 708                 xig.xig_count = udbinfo.ipi_count;
 709                 splx(s);
 710                 error = SYSCTL_OUT(req, &xig, sizeof xig);
 711         }
 712         FREE(inp_list, M_TEMP);
 713         return error;
 714 }
 715
 716 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
 717             udp_pcblist, "S,xinpcb", "List of active UDP sockets");
 718
 719
 720
 721 static int
 722 udp_output(inp, m, addr, control, p)
 723         register struct inpcb *inp;
 724         struct mbuf *m;
 725         struct sockaddr *addr;
 726         struct mbuf *control;
 727         struct proc *p;
 728 {
 729         register struct udpiphdr *ui;
 730         register int len = m->m_pkthdr.len;
 731         struct in_addr laddr;
 732         int s = 0, error = 0;
 733
 734         KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
 735
 736         if (control)
 737                 m_freem(control);               /* XXX */
 738
 739         KERNEL_DEBUG(DBG_LAYER_OUT_BEG, inp->inp_fport, inp->inp_lport,
 740                      inp->inp_laddr.s_addr, inp->inp_faddr.s_addr,
 741                      (htons((u_short)len + sizeof (struct udphdr))));
 742
 743         if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
 744                 error = EMSGSIZE;
 745                 goto release;
 746         }
 747
 748         if (addr) {
 749                 laddr = inp->inp_laddr;
 750                 if (inp->inp_faddr.s_addr != INADDR_ANY) {
 751                         error = EISCONN;
 752                         goto release;
 753                 }
 754                 /*
 755                  * Must block input while temporarily connected.
 756                  */
 757                 s = splnet();
 758                 error = in_pcbconnect(inp, addr, p);
 759                 if (error) {
 760                         splx(s);
 761                         goto release;
 762                 }
 763         } else {
 764                 if (inp->inp_faddr.s_addr == INADDR_ANY) {
 765                         error = ENOTCONN;
 766                         goto release;
 767                 }
 768         }
 769         /*
 770          * Calculate data length and get a mbuf
 771          * for UDP and IP headers.
 772          */
 773         M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
 774         if (m == 0) {
 775                 error = ENOBUFS;
 776                 if (addr)
 777                         splx(s);
 778                 goto release;
 779         }
 780
 781         /*
 782          * Fill in mbuf with extended UDP header
 783          * and addresses and length put into network format.
 784          */
 785         ui = mtod(m, struct udpiphdr *);
 786         bzero(ui->ui_x1, sizeof(ui->ui_x1));    /* XXX still needed? */
 787         ui->ui_pr = IPPROTO_UDP;
 788         ui->ui_src = inp->inp_laddr;
 789         ui->ui_dst = inp->inp_faddr;
 790         ui->ui_sport = inp->inp_lport;
 791         ui->ui_dport = inp->inp_fport;
 792         ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
 793
 794         /*
 795          * Set up checksum and output datagram.
 796          */
 797         if (udpcksum) {
 798                 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
 799                     htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
 800                 m->m_pkthdr.csum_flags = CSUM_UDP;
 801                 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
 802         } else {
 803                 ui->ui_sum = 0;
 804         }
 805         ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
 806         ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;    /* XXX */
 807         ((struct ip *)ui)->ip_tos = inp->inp_ip_tos;    /* XXX */
 808         udpstat.udps_opackets++;
 809
 810         KERNEL_DEBUG(DBG_LAYER_OUT_END, ui->ui_dport, ui->ui_sport,
 811                      ui->ui_src.s_addr, ui->ui_dst.s_addr, ui->ui_ulen);
 812
 813 #if IPSEC
 814         if (ipsec_bypass == 0 && ipsec_setsocket(m, inp->inp_socket) != 0) {
 815                 error = ENOBUFS;
 816                 goto release;
 817         }
 818 #endif /*IPSEC*/
 819         error = ip_output(m, inp->inp_options, &inp->inp_route,
 820             (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
 821             inp->inp_moptions);
 822
 823         if (addr) {
 824                 in_pcbdisconnect(inp);
 825                 inp->inp_laddr = laddr; /* XXX rehash? */
 826                 splx(s);
 827         }
 828         KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0,0,0,0);
 829         return (error);
 830
 831 release:
 832         m_freem(m);
 833         KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0,0,0,0);
 834         return (error);
 835 }
 836
 837 u_long  udp_sendspace = 9216;           /* really max datagram size */
 838                                         /* 40 1K datagrams */
 839 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
 840     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
 841
 842 u_long  udp_recvspace = 40 * (1024 +
 843 #if INET6
 844                                       sizeof(struct sockaddr_in6)
 845 #else
 846                                       sizeof(struct sockaddr_in)
 847 #endif
 848                                       );
 849 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
 850     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
 851
 852 static int
 853 udp_abort(struct socket *so)
 854 {
 855         struct inpcb *inp;
 856         int s;
 857
 858         inp = sotoinpcb(so);
 859         if (inp == 0)
 860                 return EINVAL;  /* ??? possible? panic instead? */
 861         soisdisconnected(so);
 862         s = splnet();
 863         in_pcbdetach(inp);
 864         splx(s);
 865         return 0;
 866 }
 867
 868 static int
 869 udp_attach(struct socket *so, int proto, struct proc *p)
 870 {
 871         struct inpcb *inp;
 872         int error; long s;
 873
 874         inp = sotoinpcb(so);
 875         if (inp != 0)
 876                 return EINVAL;
 877
 878         error = soreserve(so, udp_sendspace, udp_recvspace);
 879         if (error)
 880                 return error;
 881         s = splnet();
 882         error = in_pcballoc(so, &udbinfo, p);
 883         splx(s);
 884         if (error)
 885                 return error;
 886         inp = (struct inpcb *)so->so_pcb;
 887         inp->inp_vflag |= INP_IPV4;
 888         inp->inp_ip_ttl = ip_defttl;
 889         return 0;
 890 }
 891
 892 static int
 893 udp_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 894 {
 895         struct inpcb *inp;
 896         int s, error;
 897
 898         inp = sotoinpcb(so);
 899         if (inp == 0)
 900                 return EINVAL;
 901         s = splnet();
 902         error = in_pcbbind(inp, nam, p);
 903         splx(s);
 904         return error;
 905 }
 906
 907 static int
 908 udp_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
 909 {
 910         struct inpcb *inp;
 911         int s, error;
 912
 913         inp = sotoinpcb(so);
 914         if (inp == 0)
 915                 return EINVAL;
 916         if (inp->inp_faddr.s_addr != INADDR_ANY)
 917                 return EISCONN;
 918         s = splnet();
 919         error = in_pcbconnect(inp, nam, p);
 920         splx(s);
 921         if (error == 0)
 922                 soisconnected(so);
 923         return error;
 924 }
 925
 926 static int
 927 udp_detach(struct socket *so)
 928 {
 929         struct inpcb *inp;
 930         int s;
 931
 932         inp = sotoinpcb(so);
 933         if (inp == 0)
 934                 return EINVAL;
 935         s = splnet();
 936         in_pcbdetach(inp);
 937         splx(s);
 938         return 0;
 939 }
 940
 941 static int
 942 udp_disconnect(struct socket *so)
 943 {
 944         struct inpcb *inp;
 945         int s;
 946
 947         inp = sotoinpcb(so);
 948         if (inp == 0)
 949                 return EINVAL;
 950         if (inp->inp_faddr.s_addr == INADDR_ANY)
 951                 return ENOTCONN;
 952
 953         s = splnet();
 954         in_pcbdisconnect(inp);
 955         inp->inp_laddr.s_addr = INADDR_ANY;
 956         splx(s);
 957         so->so_state &= ~SS_ISCONNECTED;                /* XXX */
 958         return 0;
 959 }
 960
 961 static int
 962 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
 963             struct mbuf *control, struct proc *p)
 964 {
 965         struct inpcb *inp;
 966
 967         inp = sotoinpcb(so);
 968         if (inp == 0) {
 969                 m_freem(m);
 970                 return EINVAL;
 971         }
 972         return udp_output(inp, m, addr, control, p);
 973 }
 974
 975 int
 976 udp_shutdown(struct socket *so)
 977 {
 978         struct inpcb *inp;
 979
 980         inp = sotoinpcb(so);
 981         if (inp == 0)
 982                 return EINVAL;
 983         socantsendmore(so);
 984         return 0;
 985 }
 986
 987 struct pr_usrreqs udp_usrreqs = {
 988         udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
 989         pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
 990         pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
 991         pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
 992         in_setsockaddr, sosend, soreceive, sopoll
 993 };
 994